Drop as many PDF files as you like into the zone below, drag the cards into the order you want, and get one document back. The merge runs on your own processor: nothing is uploaded, nothing is queued, and no copy of your contract sits on a server waiting to be deleted. Free for commercial use. No watermark, and no page limit on any file you put in.
Your files are read by JavaScript running on your own machine. There is no upload, no queue, and no copy on our server to delete later.
Drag a card to reorder it, or use the arrows under each card. The order here is the order in the output. Each card is a preview of the finished page, so what you see is what the PDF will contain.
Esc close ← → move R rotate
Starting
This is your processor doing the work, so the time it takes depends on your machine and not on how many other people are on the site.
The tool above is finished work: drop the files, set the order, download the result. The guide below is the part nobody writes down - what actually happens to your documents inside a merge, why merged files get unexpectedly large, what your PDF still tells a recipient about the machine it came from, and which merges are guaranteed to fail before you waste an afternoon on them.
Two minutes, no account needed for your first run, and nothing to install. This PDF merger runs entirely inside the page you are reading, which means the four steps below happen on your computer and not on ours.
Click the drop zone above, or drag your PDFs anywhere onto this page. The whole page is the drag target rather than just the dashed rectangle, so letting go two pixels off the edge will not navigate you away and lose the batch. Each file is parsed as it lands. Its real page count appears on the card straight away.
Drag the cards to reorder them, or use the arrows on each card if you are on a phone where drag and drop is unreliable. Sort A-Z uses natural sorting, so page-2 lands before page-10 the way a human expects and not the way a plain string comparison would put it.
Strip all metadata is the default. It is also the right one: it clears the author, title and creator fields your source applications embedded, before the document ever leaves your machine, and those fields are where an old account name or a local file path usually hides. Choose Keep from first file only when the merged document is genuinely a continuation of that first file.
Press Merge PDF. The work happens on your processor, and if you left bookmarks switched on, open the result and check the navigation panel: every source file should be a named entry pointing at the page it starts on. That is the difference between a merged document and a 300 page brick.
The order shown in the workbench is the order in the output. There is no hidden alphabetical pass applied at the end, which is a behavior some combine PDF services have and never document, and which is how people end up with an appendix in front of a cover letter.
Open a PDF in a text editor and what you find is a small database. Numbered indirect objects, a cross-reference table recording which byte each one starts at, and a trailer pointing at the root of the whole thing. The pages hang off a tree there. Each page object is mostly a set of references - to its content stream, to the fonts it uses, to the images it draws, to the color spaces those images live in.
That structure is why concatenation fails. You cannot join two PDF files the way you can join two text files, because object number 14 in the second document is a completely different thing from object number 14 in the first, and every reference in both documents would now point at the wrong target. Anyone who has tried cat a.pdf b.pdf > c.pdf has watched a reader open the result and show only the first document, or refuse it entirely.
A correct merge builds a brand new document and deep-copies every page into it. For each page that means walking its resource dictionary, following every reference it holds, copying those objects across, and renumbering all of them so the references stay internally consistent in their new home. Fonts travel with their pages. A page that uses an embedded subset of a typeface brings that subset along, and a page carrying a 4000 pixel product photograph brings the image XObject with it, byte for byte, still in whatever compression it already used.
That last detail matters more than anything else about merging: the image and font data in your source files is copied, never re-encoded. A photograph that arrived as a JPEG stream inside the source PDF leaves as exactly the same JPEG stream. Nothing is decoded. With no decode and re-encode step there is no generation loss, so a scanned page in the merged document is pixel-identical to the same page in the original.
Merge an A4 report with a US Letter invoice and the output has an A4 page followed by a Letter page. Some people report that as a fault. It is not. Every page in a PDF carries its own MediaBox, and normalizing them all to one size would mean scaling somebody's signed invoice by 3 percent so it fits a page size they did not choose. If you genuinely need one uniform size, that is a resize operation applied deliberately before the merge, not something a merger should do behind your back.
Every PDF carries an information dictionary and most people have never opened one. It holds Title, Author, Subject, Keywords, Creator and Producer, plus creation and modification timestamps, and every one of those fields is filled in by software rather than by you. Word writes the account name of whoever last saved the file into Author. Scanners put their model number in Producer. Export tools frequently drop the full local file path into Title, which is how a supplier learns your internal folder structure is D:\Clients\2026\Renegotiation\FINAL_v9.
On top of that sits an XMP metadata stream, an XML block attached to the document catalog, and it is often a superset of the same information with editing history layered on. Neither survives a merge here by accident: the merged file is built from a new, empty document, so the source catalogs and their XMP streams are simply never carried across. What the info dictionary contains is then decided by the switch in the options rail, and it defaults to empty.
| Field | What usually ends up in it | After Strip all |
|---|---|---|
| Author | The operating system account name of whoever saved the file | Empty |
| Title | A local file path, or a template name nobody renamed | Empty |
| Creator | The application that authored the content | Empty |
| Producer | The library or driver that wrote the bytes | Empty |
| Keywords | Whatever a document management system injected | Empty |
| XMP stream | Editing history, tool versions, sometimes GPS from a phone | Not carried over at all |
Stripping metadata is not redaction. It cleans the document properties and leaves the pages exactly as they were. Text hidden under a black rectangle you drew in a viewer is still text underneath, still selectable, and still copy-pasteable straight out of the merged file by whoever receives it. Removing content permanently is a different operation with a different failure mode, and it is what a redaction tool is for.
Choose Keep from first file when the merged document is a continuation of that first file and the metadata is doing real work: a document management system that indexes on Title, an archival workflow with a controlled Author field, or a legal bundle where provenance is part of the point. In those cases the title, author, subject and keywords of file one are carried into the merged document and everything else is left clean.
Merge five 2 MB files and the result is rarely 10 MB. Usually more, sometimes a lot more. Two mechanisms are responsible, and knowing which one you are looking at tells you whether a compressor would help or whether the file is already as small as it is going to get.
The first is font duplication. If all five source documents embed a subset of the same corporate typeface, five separate subsets are copied into the merged file, because each one is referenced by its own pages and no merger can safely assume that two font programs with the same name are byte-identical and interchangeable. The second is per-document overhead that no longer amortizes: each source contributed its own set of structural objects, and the merged file now carries the union of all of them.
Working against that, this merge serializes the output using object streams and a compressed cross-reference stream, both introduced in PDF 1.5. Instead of writing every small object as its own uncompressed block with a plaintext byte offset in a plaintext table, non-stream objects are packed together and Flate compressed as a group, and the cross-reference table itself becomes a compressed binary stream. On a document with thousands of small objects - which is exactly what a large merge produces - that routinely removes 15 to 30 percent of the structural overhead before you have touched a single image.
The industry is quiet about this part. A merged document is dominated by its images, and the images are already compressed. The only way to make them meaningfully smaller is to decode them and re-encode them at a lower quality, or to downsample them to fewer pixels. Both are lossy, both are irreversible, and both are what "compress PDF" means on nearly every site that offers it.
So this tool has no quality slider on a merge. We refused to build one. If your merged file is too large for an email gateway, the correct fix is to reduce the source images once, on purpose, at a resolution you picked yourself, instead of letting a merger quietly degrade a document you are about to send to a paying customer.
The standard architecture for a free merge PDF online service is a file upload endpoint, a worker queue, a command line utility, and a temporary download link. It is cheap and it works. It also means your document was transmitted across the internet, written to a disk you do not control, processed by a binary you cannot inspect, and retained for some period stated in a privacy policy you did not read and cannot audit.
This tool has none of those components, because it has no server-side step at all. Your files are opened with the browser File API, parsed in memory by a PDF library executing as JavaScript on your processor, and the merged result is produced as a Blob that your browser downloads from itself. The document exists in two places. Your own disk, and your own RAM.
/wp-json/carthacker/v1/pdf/claim that carries the tool name and nothing else. No request contains your file, and no request is anywhere near its size.Think about what gets merged. A tender submission with its pricing schedule and three signed annexes. A month of supplier invoices on their way to accounts payable for reconciliation. A leaver file. A due diligence pack, or the certificates an auditor asked for by Friday. If you work somewhere that a data transfer to a third party has to be justified in writing, the honest answer here is that no transfer occurs, and that is a very different conversation from "they said they delete it after an hour".
A 300 page merged PDF with no navigation is technically a success and practically a failure. The recipient opens it, sees a scrollbar the width of a hair, and has no way to reach the fourth supplier agreement except by dragging and guessing. This is the single most common thing wrong with merged documents in the wild.
With bookmarks switched on, the merge writes a document outline: one named entry per source file, pointing at the exact page that file begins on. Every mainstream reader shows it in a side panel, and it turns a brick back into a document. Because PDF has no outline concept that survives a plain page copy, this is built directly against the document object model - an outline dictionary in the catalog, one item object per entry, each linked to its neighbors and carrying an explicit destination array that names the target page and the vertical position to scroll to.
Entry names come from the source file names with the extension removed, which is a strong argument for naming your files properly before you drop them. 2026-03-Acme-Supply-Agreement.pdf becomes a useful bookmark. scan0043.pdf becomes a useless one, and no tool can fix that for you.
page-2 precedes page-10. A plain lexicographic sort, which is what most systems do, puts page-10 first and quietly ruins a scanned book.Failures are surfaced on the file card the moment a document is read, before you spend an operation on the batch. That is deliberate: finding out mid-merge that file seven was encrypted, after the allowance was already consumed, would be an insulting way to learn it.
| What you see | What it means | What to do |
|---|---|---|
| Password protected | The document is encrypted. Its object streams cannot be read without the password, and no merger can copy pages it cannot decode. | Open it in a reader with the password, remove the protection, then merge. Our Unlock PDF tool covers this once it ships. |
| Not a readable PDF | The cross-reference table is broken or the file is not a PDF at all. Renaming a .docx to .pdf is more common than you would think. | Try opening it in any reader first. If that fails too, it needs repair rather than merging. |
| This browser cannot decode it | The file passed the extension check but the engine could not parse the object structure, usually because it was truncated by a failed download. | Download the source file again and compare the byte size. |
| The tab runs out of memory | A browser tab holds every source file and the output simultaneously. Very large batches of image-heavy scans will hit that wall. | Merge in two passes: combine halves, then combine the two results. Two operations, and it always works. |
None of these are unique to this tool. They are consequences of what a PDF is, and any merger that does not warn you about the first one is doing you a disservice.
Yes. Merge documents for a client, a supplier, an audit, or anything else that makes you money. No watermark is added to the output, no attribution is required, and there is no per-page or per-file charge. The free allowance limits how many operations you run in a 12 hour window, and it says nothing at all about what you are then permitted to do with the file you got back.
No. The merge is done by JavaScript running in your browser, on your own processor, and your documents are never transmitted anywhere. Prove it without taking our word for any of this: load this page, disconnect from the internet completely, then merge anyway. It works. There is no server-side step for it to depend on.
Twenty files and 100 MB in total per batch on the free tiers, one hundred files and 500 MB on Premium. That ceiling comes from what a browser tab can hold in memory before the operating system kills it, which is also why the figure is a batch size rather than a daily cap. Split the job. Merge the halves, then merge the two results.
No. Image data is copied out of the source documents into the merged one in its original compression, byte for byte, with no decode and re-encode step anywhere in the process. A scan in the merged file is pixel-identical to the same scan in the original. Any tool offering a quality slider on a merge is re-encoding your images, and that is lossy and irreversible.
Usually embedded fonts. If several source documents each embed a subset of the same typeface, every subset is copied, because no merger can safely assume two font programs sharing a name are interchangeable. Structural overhead from each source also adds up. The output is written with object streams and a compressed cross-reference stream, which claws back a meaningful share of that before anything else is considered.
Not while it is protected. Encryption applies to the object streams themselves, so the pages cannot be read, let alone copied, without the password. Open the file in any reader using the password, save an unprotected copy, and merge that instead. Encrypted files are flagged on their card the moment they land, before you spend an operation on the batch, rather than failing halfway through a run you have already paid for.
By default, yes. Strip all metadata is the default setting: the merged file is built as a new document, so the source XMP metadata streams are never carried over at all, and the information dictionary fields for author, title, subject, keywords, creator and producer are written empty. Choose Keep from first file if you deliberately want that document's properties preserved.
Yes, always. No tool can engineer around it, because a digital signature is a cryptographic hash of the exact bytes of the document it was applied to, and merging produces a different document with different bytes. The signature stops validating. Merge first, then sign the combined document.
They are generally flattened in practice. Field definitions live in a document-level AcroForm dictionary rather than on the pages, so copying pages into a new document brings the visual appearance of the fields across without the machinery that makes them fillable. If the fields have to stay live, keep those documents separate or fill them before merging.
Yes, and each page keeps its own dimensions in the output. That is correct behavior rather than a fault: every PDF page carries its own MediaBox, and forcing them all to one size would mean silently scaling somebody's signed invoice to fit a page format they never chose.
Yes. Drag the cards into any order, or use the arrow buttons on each card, which also work on touch screens and with a keyboard. Sort A-Z applies a natural sort that reads embedded numbers as numbers, so page-2 comes before page-10 rather than after it, which is where a plain alphabetical sort would put it. The workbench order is the output order.
Not for your first operation. Anonymous visitors get one operation per 12 hours counted by network address, a free account gets two, and Premium removes the counter. The limit exists because running the tool costs us nothing per operation, so the only thing worth metering is the audience, and we would rather say that plainly than pretend it is a technical constraint.
Yes. The interface is built for touch as well as pointer input, which is why every file card carries explicit move buttons instead of relying on drag and drop, since HTML5 drag events are unreliable on mobile browsers. The practical limit on a phone is memory: large batches of high-resolution scans are far more likely to exhaust a mobile tab than a desktop one.
A desktop editor is a licensed application with authoring, redaction, preflight and OCR, and it is worth paying for if you live in documents all day. This is a single operation done properly, for free, with no installation and no upload. Use the editor for production document work. Use this when you need three invoices to become one file before accounting will accept them.
All of these run the same way: your file is read by JavaScript on your own machine, the work happens on your processor, and the result is handed back to you as a download that never existed anywhere else.
Find the apps that slow your store down, by name and in milliseconds, then answer shoppers' doubts with trust badges under Add to cart. Both are free, with no card and no access to your store admin.
Paste any page and get its three Core Web Vitals, the apps costing time, like Klaviyo or Judge.me, and the Shopify or WooCommerce fix.
Show the promises that settle doubts, from free shipping to a money-back guarantee, right under the Add to cart button. One line in your theme.
Change the link after printing and count the scans. 5 tracked codes on a free account, static codes unlimited.
Anonymous visitors get one operation per 12 hours, per network address. That is not a sales tactic, it is the only lever we have: this tool runs on your processor, not ours, so the run itself earns us nothing. An account tells us who you are, and the ceiling goes up the moment you have one. Your allowance resets in 12 h.
No credit card, no trial period, no sales call. Join with Google, GitHub, Facebook or email.
Free accounts get two operations per 12 hours, and yours reset in 12 h. Waiting costs you nothing and is a completely reasonable choice. If PDF work is a weekly job rather than a one-off, the counter is the wrong tool for you and Premium takes it off the page.
Cancel anytime. Everything you have already produced stays yours either way.